Chapter 6 Noncommutative Spaces Fedele Lizzi In this chapter we present some of the basic concepts needed to describe noncommu- tative spaces and their topological and geometrical features. We therefore comple- ment the previous chapters where noncommutative spaces have been described by the commutation relations of their coordinates. The full algebraic description of or- dinary (commutative) spaces requires the completion of the algebra of coordinates into a C⋆-algebra, this encodes the Hausdorff topology of the space. The smooth manifold structure is next encoded in a subalgebra (of “smooth” functions). Relax- ing the requirement of commutativity of the algebra opens the way to the definition of noncommutative spaces, which in some cases can be a deformation of an ordinary space. A powerful method to study these noncommutative algebras is to represent them as operators on a Hilbert space. We discuss the noncommutative space gener- ated by two noncommuting variables with a constant commutator. This is the space of the noncommutative field theories described in this book, as well as the elemen- tary phase space of quantum mechanics. The Weyl map from operators to functions is introduced in order to produce a ⋆-product description of this noncommutative space. 6.1 Commutative geometry (and topology) In Hilbert’s foundations of geometry [1] the concepts of points, lines, and planes are considered intuitive and no attempt is made to define them. These “undefined” points are nevertheless the basis of any topological space, differentiable manifold, bundle, and so on, all geometrical concepts built on spaces made of points. This gave the impression that geometrical notions cannot survive without points. Quan- tum mechanics forced a change of this attitude. While in classical mechanics the state of a system can be described by a point in a phase space, Heisenberg’s uncer- tainty principle makes the concept impossible in quantum mechanics. This led von Neumann [2] to speak of pointless geometry. Lizzi, F.: Noncommutative Spaces. Lect. Notes Phys. 774, 89–109 (2009) DOI 10.1007/978-3-540-89793-4 6 c Springer-Verlag Berlin Heidelberg 2009 90 Fedele Lizzi In the following we introduce the basic mathematics of noncommutative geom- etry at an unspecialized level, that of a high-energy physics student for example. Sometimes we sacrifice rigor and refer to some of the classic reference books [3–6] for details and proofs. An extended and more rigorous treatment of the topics of this chapter will also appear in the forthcoming book [7]. In the present section we discuss ordinary topology and geometry from a point of view which enables its generalization to a noncommutative setting. The main tool is the transcription of the usual geometrical concepts in terms of algebras of operators. The starting point is a series of theorems due to Gel’fand and Naimark (for a review see for example [4, 5]). They established a complete equivalence between Hausdorff topological spaces and commutative C∗-algebras. From a physicist point of view one can look at this activity as describing the topology (and geometrical properties) of a space not seeing it as a set of points, but as the set of fields defined on it. In this sense the tools of noncommutative geometry resemble the methods of modern theoretical physics. 6.1.1 Topology and algebras A topological space M is a set on which a topology is defined: a collection of open subsets obeying certain conditions, this enables the concept of convergence of suc- cession of points xn M to a limit point x = limn xn. Together with the concept of convergence goes the∈notion of continuous function. A function from a topological space into another topological space is continuous if the inverse image of an open set is open, but as a consequence it maps convergent sequences into convergent se- quences: lim f (xn)= f (x). (6.1) n AHausdorff topology makes the space separable, i.e., given two points it is always possible to find two disjoint open sets each containing one of the two points. The common topological spaces encountered in physics (for example, manifolds) are separable. Of particular interest in this context is the set of complex-valued continuous func- tion. They form a commutative algebra because the sum or product of two continu- ous functions is still continuous. We will show how it is possible to define the topol- ogy of a space from the algebra of continuous functions on it. Moreover, we will show how to construct the topological space starting from the abstract algebra. On one hand every Hausdorff topological space defines naturally a commutative alge- bra, the algebra of continuous complex-valued functions over it. Remarkably, under certain technical assumptions spelled below, the reverse is also true, i.e., given a commutative algebra A as an abstract entity, it is always possible to find a topologi- cal space whose algebra of continuous functions is A . Therefore, we can establish a complete equivalence between topological spaces and algebras. In the following we 6 Noncommutative Spaces 91 will describe these mathematical structures from an “user” point of view, keeping the technicalities at a minimum and refer the literature for proofs and details. The technical assumptions we have mentioned are resumed in the fact that the algebra A must be a C∗-algebra. This is, first of all, a vector space with the structure of an associative algebra over the complex numbers C, i.e., a set on which we can define two operations, sum (associative and commutative) and product (associative but not necessarily commutative), and the product of a vector by a complex number, with the following properties: (1) A is a vector space over C, i.e., αa + βb A for a,b A and α,β C. (2) It is distributive over addition with respect∈ to left and ∈right multiplication,∈ i.e., a(b + c)=ab + ac and (a + b)c = ac + bc, a,b,c A . ∀ ∈ A is further required to be a Banach algebra: (3) It has a norm : A R with the usual properties · → a) a 0 , a = 0 a = 0 b) α≥a = α a ⇐⇒ c) a +b | |a + b d) ab ≤a b ≤ The Banach algebra A is called a -algebra if, in addition to the properties above, it has a hermitian conjugation operation∗ (analogous to the complex conjugation defined for C) with the properties ∗ (4) (a∗)∗ = a (5) (ab)∗ = b∗a∗ ¯ (6) (αa + βb)∗ = α¯ a∗ + βb∗ (7) a = a ∗ (8) a a = a 2 ∗ for any a,b A and α,β C, where α¯ denotes the usual complex conjugate of α C. Finally∈ , ∈ ∈ (9) It is complete with respect to the norm. C∗-algebras play a very important role in mathematics because as we will see their study is basically the study of topology. A good introduction to their properties is found in the book [8]. Example 6.1. Examples of C∗-algebras are n n matrices, bounded operators on an infinite- dimensional Hilbert space, as well× as compact operators. The norm is the supre- mum norm in all these cases. These are noncommutative, examples of commutative algebras are C itself, or the continuous functions on the plane. Note that several commonly used algebras do not satisfy all of the definitions. For example, the set of upper triangular matrices does not have the hermitian conjugation, trace class operators are not complete, and the Hilbert space of L2 functions has a norm which does not satisfy item (8) above. 92 Fedele Lizzi Every Hausdorff topological space has a natural commutative C∗-algebra asso- ciated with it: the algebra of continuous complex-valued functions. If the space is compact this algebra contains the unity I and is called unital. The converse is also true. Every unital commutative C∗-algebra is the C∗-algebra of continuous func- tions on some compact topological space. Nonunital algebras are similarly associ- ated with noncompact Hausdorff spaces. 6.1.2 Reconstructing the space from the algebra We now show how the topological space can be reconstructed from the algebra. We first introduce the notion of state. A state is a linear functional from a C∗-algebra A (not necessarily commutative) into complex numbers: φ : A C, (6.2) −→ with the positivity and normalization requirements φ(a∗a) 0 a A , φ = 1 . (6.3) ≥ ∀ ∈ In this case the norm is defined as φ = sup φ(a) . (6.4) a 1{ } ≤ If the algebra is unital φ(I)=1. The space of states is convex, i.e., any linear combination of states of the kind 2 2 cos λφ1 + sin λφ2 is still a state for any value of λ. Some states cannot be ex- pressed as such convex sum, they form the boundary of the set and are called pure states. Example 6.2. Consider the case of n n complex-valued matrices. A state is given by a matrix (which with an abuse we×still call φ) with the definition φ(a)=Trφa. (6.5) Positivity requires the matrix φ to be self-adjoint with positive eigenvalues, and normalization requires it to have unit trace. Since the matrix is self-adjoint it can be diagonalized. There are two possibilities. Either more than one eigenvalues is different from zero, and in this case it is immediate to see that we can write it as the convex sum of two diagonal matrices of trace 1. Alternatively only one eigenvalue is different from zero, and it must be the unity. In this case it is not possible to express φ as the convex sum of two matrices of trace 1, since positivity requires diagonal elements to be positive numbers less than 1.